The Effects of Volume Bid Contracts on Service Level Agreements

Introduction

This project assesses the impact of implementing volume bid contracts on service-level agreements governing the maintenance of data center equipment. Specifically, this study investigates whether the use of volume bid contracts negatively influences the average equipment repair times in data centers as compared to traditional contracting approaches.

Table of Contents

• Introduction
• Section 1 - Set Environment
• Section 2 - Load and Process Data
• Section 3 - Visual Analysis
• Section 4 - Outlier Analysis
• Section 5 - Data Distribution Analysis
• Section 6 - Data Transformation Analysis
• Section 7 - Independent T-Test Analysis
• Section 8 - Mann-Whitney U Test Analysis
• Section 9 - Parametric and Nonparametric Analysis
• Section 10 - Chi-Square Analysis
• Conclusion

Section 1 - Set Environment

Purpose

The purpose of this section is to load the necessary Python packages, establish a path to the application directory, and initialize the Analysis Main class object.

Section Contents

1. Import Packages and Libraries
2. Set Path to Python Files
3. Create the Analysis Main Object

Set Environment 1 - Import Packages and Libraries

import os
import sys

Set Environment 2 - Set Path to Python Files

module_path = os.path.abspath(os.path.join('..', 'Python'))
if module_path not in sys.path: sys.path.append(module_path) 

Set Environment 3 - Create the Analysis Main Object

from data_analysis_main import AnalysisMain
am = AnalysisMain()

Section 2 - Load and Process Data

Purpose

This section focuses on uploading individual data files and integrating them into a comprehensive dataset, as well as conducting feature engineering to facilitate thorough data analysis.

This section aligns with the following:

• Objective 1.1 of the Project Plan

Section Contents

1. Load Data
2. Process Data

Load and Process Data 1 - Load Data from CSV Files

am.load_data()

Load and Process Data 2 - Process Loaded Data

am.process_data()

Observations

Following the successful completion of the Data Collection, Processing, and Feature Engineering phases, the subsequent observations were documented.

• The historical dataset contained a total of 28,385 unique entries.
• The current dataset contained a total of 36,906 unique entries.
• The historical dataset contained 90 unique vendors.
• The current dataset contained 40 unique vendors.
• The processed dataset contains 65,291 unique entries.
• The processed dataset contains 13 original features and 10 engineered features.

Section 3 - Visual Analysis

Purpose

The objective of this section is to conduct a thorough visual examination of the data using histograms, box plots, bar plots, violin plots, and KDE plots. Furthermore, it encompasses performing descriptive statistical analyses and presenting the findings in a tabular format.

This section aligns with the following:

• Objective 1.2 of the Project Plan
• Section 1 of the Data Analysis Process

Section Contents

1. Histogram Plot Analysis
2. Box Plot Analysis
3. Count Plot Analysis of Priority Levels
4. Violin Plot Analysis of Hours to Complete by Priority Levels
5. Violin Plot Analysis of Days Past Due by Priority Levels
6. Count Plot Analysis of SLA Label
7. KDE Plot Analysis of SLA Label
8. Statistical Summary of Hours to Complete by SLA Label
9. Statistical Summary of Days Past Due by SLA Label

Visual Analysis 1 - Histogram Plot Analysis of Data

am.histogram_plot_analysis()

Visual Analysis 2 - Box Plot Analysis of Data

am.box_plot_analysis()

Visual Analysis 3 - Count Plot Analysis of Priority Levels

am.count_plot_analysis_1()

Visual Analysis 4 - Violin Plot Analysis of Hours to Complete by Priority Levels

am.violin_plot_analysis_1()

Visual Analysis 5 - Violin Plot Analysis of Days Past Due by Priority Levels

am.violin_plot_analysis_2()

Visual Analysis 6 - Count Plot Analysis of SLA Label

am.count_plot_analysis_2()

Visual Analysis 7 - KDE Plot Analysis of SLA Label

am.kde_plot_analysis()

Visual Analysis 8 - Statistical Summary of Hours to Complete by SLA Label

am.statistic_summary_1()

Visual Analysis 9 - Statistical Summary of Days Past Due by SLA Label

am.statistic_summary_2()

Observations

Following the successful completion of the visual and statistical analysis, the subsequent observations were documented.

• The means of the historical data are significantly lower than those of the current data.
• The ‘Hours to Complete’ attribute is substantially right-skewed in both datasets.
• The historic data exhibits less dispersion and fewer outliers compared to the current data. P2 and P6 are the priority levels with the most significant number of entries in both datasets.
• The historical data met the SLA target 90.7% of the time compared to only 51.6% for the current data.
• The means for both SLA Met and SLA Not Met were noticeably lower in the historical data than in the current data.

Section 4 - Outlier Analysis

Purpose

The objective of this section entails a thorough analysis of the dataset to identify outliers and potential anomalies. The findings are presented using box plots, histograms, and comprehensive statistical tables.

This section aligns with the following:

• Objective 1.3 of the Project Plan
• Section 2 of the Data Analysis Process

Section Contents

1. Analysis of Hours to Complete for the Current Data
2. Analysis of Hours to Complete for the Historic Data
3. Analysis of Days Past Due for the Current Data
4. Analysis of Days Past Due for the Historic Data

Outlier Analysis 1 - Analysis of Hours to Complete for the Current Data

am.outlier_analysis_1()

Outlier Analysis 2 - Analysis of Hours to Complete for the Historic Data

am.outlier_analysis_2()

Outlier Analysis 3 - Analysis of Days Past Due for the Current Data

am.outlier_analysis_3()

Outlier Analysis 4 - Analysis of Days Past Due for the Historic Data

am.outlier_analysis_4()

Observations

Following the successful completion of the data outlier analysis, the subsequent observations were documented.

• The total number of outliers was significantly lower in the historical data compared to the current data.
• The current data has 4255 more outliers in the ‘Days Past Due’ attribute than in the ‘Hours to Complete’ attribute.
• The historic data has 207 more outliers in the ‘Hours to Complete’ attribute than in the ‘Days Past Due’ attribute.
• The confidence interval is considerably smaller in the historical data than in the current data for both ‘Hours to Complete’ and ‘Days Past Due’ attributes.

Section 5 - Data Distribution Analysis

Purpose

The objective of this section is to conduct a comprehensive distribution analysis of the dataset to assess its normality, employing both the Shapiro-Wilk and Anderson-Darling tests. The evaluation additionally encompasses an examination of skewness and kurtosis.

This section aligns with the following:

• Objective 1.4 of the Project Plan
• Section 3 of the Data Analysis Process

Section Contents

1. Analysis of Hours to Complete for the Current Data
2. Analysis of Hours to Complete for the Historic Data
3. Analysis of Days Past Due for the Current Data
4. Analysis of Days Past Due for the Historic Data

Distribution Analysis 1 - Analysis of Hours to Complete for the Current Data

am.distribution_analysis_1()

Distribution Analysis 2 - Analysis of Hours to Complete for the Historic Data

am.distribution_analysis_2()

Distribution Analysis 3 - Analysis of Days Past Due for the Current Data

am.distribution_analysis_3()

Distribution Analysis 4 - Analysis of Days Past Due for the Historic Data

am.distribution_analysis_4()

Observations

Following the successful completion of the data distribution analysis, the subsequent observations were documented.

• The current and historical data are non-normally distributed for both ‘Hours to Complete’ and ‘Days Past Due’ attributes.
• The p-value for the Shapiro-Wilk test was less than 0.000000 for both current and historical datasets in the ‘Hours to Complete’ and ‘Days Past Due’ attributes.
• The p-value for the Anderson-Darling test was less than 0.000000 for the current and historical datasets in both ‘Hours to Complete’ and ‘Days Past Due’ attributes.
• Both the current and historical datasets are positively skewed for the ‘Hours to Complete’ attribute.
• Both the current and historical datasets are negatively skewed for the ‘Days Past Due’ attribute.
• The data distribution has heavier tails than a normal distribution for both current and historical datasets in the ‘Hours to Complete’ and ‘Days Past Due’ attributes.

Section 6 - Data Transformation Analysis

Purpose

The objective of this section is to execute a comprehensive transformational analysis on the dataset by applying five established transformation techniques to determine the most appropriate methodology for the data. The evaluation results are presented through histograms, Q-Q plots, and statistical tables.

This section aligns with the following:

• Objective 1.5 of the Project Plan
• Section 4 of the Data Analysis Process

Section Contents

1. Analysis of Hours to Complete for the Current Data
2. Analysis of Hours to Complete for the Historic Data
3. Analysis of Days Past Due for the Current Data
4. Analysis of Days Past Due for the Historic Data

Transformation Analysis 1 - Analysis of Hours to Complete for the Current Data

am.transformation_analysis_1()

Transformation Analysis 2 - Analysis of Hours to Complete for the Historic Data

am.transformation_analysis_2()

Transformation Analysis 3 - Analysis of Days Past Due for the Current Data

am.transformation_analysis_3()

Transformation Analysis 4 - Analysis of Days Past Due for the Historic Data

am.transformation_analysis_4()

Observations

Following the successful completion of the transformational analysis, the subsequent observations were documented.

• The Yeo-Johnson method produced the best results in transforming the current data with the ‘Hours to Complete’ attribute.
• The Box-Cox method yielded the best results in transforming the historic data with the ‘Hours to Complete’ attribute.
• The Box-Cox method prompted the best results in transforming the current data with the ‘Days Past Due’ attribute.
• The Box-Cox method generated the best results in transforming the historical data with the ‘Days Past Due’ attribute.
• The ‘Hours to Complete’ attribute transformed more effectively than the ‘Days Past Due' attribute.

Section 7 - Independent T-Test Analysis

Purpose

The objective of this section involves implementing Independent Samples T-Tests on the dataset, assessing statistical significance by comparing current data to historical data. The findings were presented via box plots, violin plots, histograms, and tabular summaries.

This section aligns with the following:

• Objective 1.6 of the Project Plan
• Section 5 of the Data Analysis Process

Section Contents

1. Analysis of Hours to Complete with the Original Data Format
2. Analysis of Days Past Due with the Original Data Format
3. Analysis of Hours to Complete with the Transformed Data Format
4. Analysis of Days Past Due for the Historic Data

Independent T-Test Analysis 1 - Analysis of Hours to Complete with the Original Data Format

am.independent_ttest_analysis_1()

Independent T-Test Analysis 2 - Analysis of Days Past Due with the Original Data Format

am.independent_ttest_analysis_2()

Independent T-Test Analysis 3 - Analysis of Hours to Complete with the Transformed Data Format

am.independent_ttest_analysis_3()

Independent T-Test Analysis 4 - Analysis of Days Past Due with the Transformed Data Format

am.independent_ttest_analysis_4()

Observations

Following the successful completion of the Independent Samples T-Test analysis, the subsequent observations were documented.

• The p-value for the T-Tests was less than 0.000000 for both original and transformed versions of the data in the ‘Hours to Complete’ and ‘Days Past Due’ attributes when comparing the current and historical datasets.
• The effect size, Cohen’s d, is significant for both the ‘Hours to Complete’ attribute and the ‘Days Past Due’ attribute.
• There is a statistically significant difference in both original and transformed versions of the data for the ‘Hours to Complete’ and ‘Days Past Due’ attributes when comparing the current and historical datasets.
• These results justify the rejection of the null hypothesis in favor of the alternative hypothesis.

Section 8 - Mann-Whitney U Test Analysis

Purpose

The objective of this section involves conducting a comprehensive non-parametric analysis of the data utilizing the Mann-Whitney U Test to evaluate statistical differences between current and historical datasets. The results are illustrated through box plots, violin plots, histograms, ECDF plots, and tabular summaries.

This section aligns with the following:

• Objective 1.7 of the Project Plan
• Section 6 of the Data Analysis Process

Section Contents

1. Analysis of Hours to Complete
2. Analysis of Days Past Due

Mann-Whitney U Test Analysis 1 - Analysis of Hours to Complete

am.mannwhitney_utest_analysis_1()

Mann-Whitney U Test Analysis 2 - Analysis of Days Past Due

am.mannwhitney_utest_analysis_2()

Observations

Following the successful completion of the Mann-Whitney U Test analysis, the subsequent observations were documented.

• The p-value for the U-Tests was less than 0.000000 for both the ‘Hours to Complete’ and ‘Days Past Due’ attributes when comparing the current and historical datasets.
• The effect size R is significant for the ‘Hours to Complete’ attribute; however, it is relatively small for the ‘Days Past Due’ attribute.
• There is a statistically significant difference in both the ‘Hours to Complete’ and ‘Days Past Due’ attributes when comparing the current and historical datasets.
• These results justify the rejection of the null hypothesis in favor of the alternative hypothesis.

Section 9 - Parametric and Nonparametric Analysis

Purpose

The objective of this section is to evaluate the effectiveness of parametric and non-parametric tests applied to the dataset, to determine the most appropriate analytical approach. Results are systematically presented in tabulated form.

This section aligns with the following:

• Objective 1.8 of the Project Plan
• Section 7 of the Data Analysis Process

Section Contents

1. Analysis of Hours to Complete
2. Analysis of Days Past Due

Parametric and Nonparametric Analysis 1 - Analysis of Hours to Complete

am.parametric_nonparametric_analysis_1()

Parametric and Nonparametric Analysis 2 - Analysis of Days Past Due

am.parametric_nonparametric_analysis_2()

Observations

Following the successful completion of the parametric and non-parametric analysis, the subsequent observations were documented.

• The current and historical data are non-normally distributed for both ‘Hours to Complete’ and ‘Days Past Due’ attributes.
• The p-value for the T-Tests was less than 0.000000 with the original version of the data for both the ‘Hours to Complete’ and ‘Days Past Due’ attributes when comparing the current and historical datasets.
• The p-value for the T-Tests was less than 0.000000 with the transformed versions of the data for both the ‘Hours to Complete’ and ‘Days Past Due’ attributes when comparing the current and historical datasets.
• The p-value for the U-Tests was less than 0.000000 for both the ‘Hours to Complete’ and ‘Days Past Due’ attributes when comparing the current and historical datasets.
• All tests yielded similar results regarding statistical significance.
• These findings conclusively support the rejection of the null hypothesis in favor of the alternative hypothesis.

Section 10 - Chi-Square Analysis

Purpose

The objective of this section is to conduct a Chi-Squared Analysis of the dataset to determine whether there are statistically significant deviations between observed and expected outcomes. The analysis results are presented using a heatmap visualization and summarized in comprehensive statistical tables.

This section aligns with the following:

• Objective 1.9 of the Project Plan
• Section 8 of the Data Analysis Process

Section Contents

1. Analysis of Dataset Name and SLA Label

Chi-Square Analysis 1 - Analysis of Dataset Name and SLA Label

am.chi_square_test_analysis_1()

Observations

Following the successful completion of the Chi-Squared analysis, the subsequent observations were documented.

• The historical data met the SLA target 90.7% of the time compared to only 51.6% for the current data.
• The p-value from the test was less than 0.000000, which is substantially lower than the pre-established alpha of 0.05.
• The findings provide sufficient evidence to reject the null hypothesis in favor of the alternative hypothesis.
• The analysis concluded that a statistically significant association exists between the ‘data-name’ and the ‘SLA-label’ attributes.

Conclusion

This project aimed to evaluate the impact of implementing volume bid contracts on service-level agreements that oversee the maintenance of data center equipment. Specifically, this study examined whether the use of volume bid contracts adversely affected the average repair times of equipment in data centers compared with conventional contracting methods.

The study demonstrated that a statistically significant difference exists in the means of both attributes when comparing current and historical contract data. This conclusion was reached by applying the Independent Samples T-Test to both the original and transformed data sets, and the Mann-Whitney U test to the original data from both contract models. All tests yielded probability values approximating zero, indicating that the observed differences are unlikely due to random variation.

An additional examination of the data, through the application of chi-squared analysis, demonstrated that there was a statistically significant association between current and historical data and the ‘SLA label’ attribute. The analysis results indicated that the current contract model for maintenance tasks achieved their SLA objectives only 51.6% of the time, in contrast to the 90.7% success rate observed in the historical contract model.

From a pragmatic standpoint, recognizing the existence of such a discrepancy indicates an issue affecting the overall effectiveness of the current contractual strategy. The observed decline in efficiency suggests a diminished ability among service providers to fulfill the organization's service level agreements.